Backscattering values across multi temporal S1 images

I stacked 4 S1 GRD data but i find out, at the same location, the dB values ranging from dB to dB. Confusingly, acrosss these 4 data, there are 2 pairs of data that are closer to each other in terms of back scattering values:

Sigma0_VH_09May2017_mst_db: -8.51345 intensity_db
Sigma0_VH_27May2017_slv1_db: -19.66345 intensity_db
Sigma0_VH_14June2017_slv2_db: -7.75498 intensity_db
Sigma0_VH_02July2017_slv3_db: -21.21673 intensity_db

The pixel data that I extracted the values is a forest area. How can these data differ so much in terms of dB values and which one that I can trust?

Soil moisture and canopy moisture influence the backscatter a great deal. In other words, that is normal and you can trust all of the values.

Can you share with us your processing steps?

I am also interested in your processing steps.
What is the best way to perform a reliable change detection on Sentinel 1 product?
I see there is a change detection in SNAP menu, how to interpret it? can anyone point to some litteratures I can read?

Can it be due to speckle noise?
Instead of informing single pixel values, it would be better to trust in mean values of an homogeneous area.

Hi, Yes certainly using a homogeneous area will give a much more representative radar cross-section than using an individual pixel (due as you say to speckle). You should then obtain similar RCS values from the four date.